Only honor second-line encoding comments after blank or comment lines - #1465

Fixes #1463

Summary

babel.util.parse_encoding() previously ran ast.parse() on line 1 to decide whether line 2 could contain a magic encoding comment. Single-line executable code (such as x = 1, """docstring""", or ;) succeeded under ast.parse(), causing Babel to honor a second-line # coding: comment that Python itself ignores per PEP 263 (and to raise a spurious SyntaxError if the file also began with a UTF-8 BOM).

This change replaces the ast.parse() check with _PYTHON_BLANK_OR_COMMENT_re (rb'[ \t\f]*(?:[#\r\n]|$)', matching tokenize.blank_re), so line 2 is checked only when line 1 is blank or a comment.

Changes

• babel/util.py: Match line 1 against _PYTHON_BLANK_OR_COMMENT_re before reading line 2 for PYTHON_MAGIC_COMMENT_re in parse_encoding().
• tests/test_util.py: Add unit tests covering valid blank/shebang/comment first lines, ignored second-line declarations after executable code/docstrings/;/\x0b, UTF-8 BOM + code on line 1, and stream position preservation.
• tests/messages/test_extract_python.py: Add extraction tests verifying second-line # coding: behavior after code, blank/comment lines, and a UTF-8 BOM.

Verification

• python -X utf8 -m pytest tests/test_util.py tests/messages/test_extract.py tests/messages/test_extract_python.py: 67 passed
• python -X utf8 -m pytest (with imported CLDR data): 8225 passed, 5 skipped, 2 xfailed
• ruff check .: all checks passed